Pulse oximetry is noninvasive, fast, and simple, making it a very popular way of assessing oxygenation in pediatric patients. However, there are few studies that establish the accuracy of this technology over a wide range of oxygen saturations in children. This study, done in 47 children aged from 1 day to 16 years with congenital heart disease and undergoing cardiac catheterization, compared the direct measurement of arterial oxygen saturation to values from pulse oximetry. Oxygen saturation was measured by an IL-282 Co-oximeter, which also measured carboxyhemoglobin and methemoglobin, and was compared to values obtained from both a Biox III and Nellcor N100. Both pusle oximeters gave values that closely correlated with the actual saturation (r = 0.91 and 0.93, respectively) with standard errors of the estimate of 4.1 and 3.2%, respectively. For both devices, the error increased with decreasing saturations, being progressively larger below a saturation of 80%. The difference between the actual saturation and that measured by pulse oximetry bore no relationship to the presence of carboxyhemoglobin, methemoglobin, fetal hemoglobin, bilirubin, cardiac index, or age of the patient. In conclusion, pulse oximetry, while a very useful technology in pediatrics, must be interpreted with some caution in children with severe cyanosis.